Implement Custom Language Models AI-102 Practice Questions (Page 15)

Questions 71–73

1. 71application · medium

   A team has built a custom question answering knowledge base for a telecom company's support site. Before publishing, they want to verify that the knowledge base returns correct answers for common questions. What should they do?

   • APublish the knowledge base and ask a few employees to test it in production.
   • BExport the knowledge base to Excel and manually review all Q&A pairs.
   • CTrain a custom language model with the same questions to compare accuracy.
   • DUse the Test pane in the Language Studio to submit sample questions and review the responses and confidence scores.
   Reveal Solution Discussion Select an answer first
2. 72foundation · easy

   Which of the following is a valid reason to export a knowledge base?

   • ATo increase the number of Q&A pairs
   • BTo change the language of the knowledge base
   • CTo migrate it to a different Azure region
   • DTo test the knowledge base
   Reveal Solution Discussion Select an answer first
3. 73application · medium

   A custom language model has been deployed to production. The support team reports that the model sometimes confuses the 'CancelOrder' and 'ReturnOrder' intents. The team wants to verify the issue and decide whether to retrain. What should they do first?

   • AImmediately retrain the model with more utterances for both intents.
   • BDelete the deployed model and create a new one from scratch.
   • CCheck the training data to see if the two intents have similar utterances.
   • DSend a set of representative queries to the deployed endpoint and review the predicted intents.
   Reveal Solution Discussion Select an answer first